全志v831 和 v833 对比表格
mangogeek last edited by mangogeek
今天, 我发现, V831的LCD和RMII是可以分开的!
Please consider using multiple I2S mics on dev board to support beamforming.
I am also curious to know if the SDK has a webcam mode. A mode that turns V831 dev board into a UVC USB web camera. A mode like that could be extended to allow control of the AI unit via a separate endpoint.
我也很想知道SDK是否具有网络摄像头模式。 一种将V831开发板变成UVC USB网络摄像头的模式。 可以扩展类似的模式以允许通过单独的端点控制AI单元。
@jonsmirl It only has one I2S , We can design two I2S MIC
The I2S supports TDM mode.
In this way, how many mics should be supported at most?
It can directly connect to an analog mic
There is no real need for more than four mics on the board. Mount them as far apart as possible. You can attach up to eight but it doesn't make a lot of difference after the first four. You will need to shop around and see if you can locate a reasonably priced I2S mic with TDM support. Not sure if you can locate a cheap one.
If you want to get extra fancy you can put those serially chained LEDs from Christmas lights by each mic so that you have a direction indicator.
If you want something extra to sell, you could leave the mics off from the base board and then make a round PCB with 8 mics and 8 LEDs. Sell it for an added charge like the K210 board do.
If you can't locate a cheap I2S mic with TDM, two mics will work for basic usage.
two i2s mic
I am using it and it works great
Two mics will not be very good at removing a television playing in the same room. He is probably using it somewhere without external noise sources.
Is three mics ok? like: mic1 ---- mic2 ---- mic3
In fact r328 or K210 is more suitable for mic array
Two or four mic is not a critical feature. Someone still has to write the software to make it work. Maybe you can borrow that code from the R328 SDK? Two is probably fine for a dev board.
Three mics will work, but it is not going to work if two of them I2S and the third is analog. The algorithm only works if all of the mics are sampled at the exact same instant. The TDM mics should all sample synchronously on the word clock and then read out the sampled value serially via the TDM time slots.
The more mics there are, the more accurately you can calculate the angle to the speaker. And the more accurately you know that angle the more you can discard not arriving on that angle. But two mics are fine for a dev board. If someone want to make a production device they can put four on their board.
Thanks, I understand
I will try to let the hardware support the follow-up design.
For the software part, I will refer to the R328 SDK
I have been looking at V831 more and I am convinced a V831 board will beat a K210 one.
- H.264 encode. This is a major thing missing from K210, there is no way to record the video the AI unit sees so that you can use it later for training.
- Easier to program, it runs Linux easily. K210 Linux hard.
- More memory, K210 memory quite constrained
- SOC is slightly cheaper.
- ISP for video noise reduction and resizing
- Common ARM CPU with compilers available everywhere
- OTG USB
- Much better audio support on K210. Allwinner could have done more on V831.
- No FFT unit
I'd copy the K210 Maix product offerings, they have a good array of choices.
Another mistake Allwinner made is no h.264 decode. That prevents you from making a sports cam type device with this chip. There is no way to display the saved h.264 video.
V833 has 2 i2s , Where i2s1 has two data lane.
four i2s mic ? 2L +2R ?
Those two I2S lanes on V833 likely for four mics. TDM I2S mic more expensive than non-TDM ones so with four lanes you can use the cheaper versions. I don't know any other reason for two I2S lanes other than to support 4 mics.
Two mics fine on V831 board.
Those pins also useful for attaching smart mic chips.